Data Scientist Roles and Skills

Context: FIT1043_MOC · who does what in a data team · the skill mix behind Conway’s diagram · deepens Data Science vs Related Disciplines (source: Analyzing the Analyzers, 2013)

Quick Revision

• 🎯 Objective: separate the three data roles by their primary output ➔ analyst = insights, scientist = models/products, engineer = infrastructure.
• ⚡ Key Constraint: “primarily” — the roles overlap; a data scientist is better at stats than a software engineer, and better at software engineering than a statistician.

📝 Core

1. The Three Roles (Jason Widjaja)

• Data analyst ➔ primarily develops insights with data.
• Data scientist ➔ primarily develops data models and products that in turn produce insights.
• Data engineer ➔ primarily manages data infrastructure, automates processing, deploys models at scale.

2. Five Skill Groups

• Business ➔ product development, business.
• ML / Big data ➔ structured + unstructured data, machine learning, big/distributed data.
• Maths / Operations research ➔ optimisation, mathematics, graphical models, algorithms.
• Programming ➔ sysadmin, back-end, front-end.
• Statistics ➔ visualisation, temporal/spatial stats, surveys/marketing, science, data manipulation.

3. Four Styles + Handbook Lessons

• Four styles ➔ Data Businesspeople, Data Creatives, Data Developers, Data Researchers.
• Communication is underrated ➔ a top, under-valued skill; keep curiosity about data.
• Hardest steps ➔ the biggest challenge for an analyst is Collection and Wrangling (see Data Science Process (Standard Value Chain)).
• CV to specialise ➔ solid ML + statistics, related maths, prototyping (R/Python/Java, GitHub, Kaggle), Unix experience.

⚠️ Common Mistakes

• 💡 “Primarily”, not “only” ➔ the three roles overlap heavily; the industry is nascent, so titles/boundaries are fluid.
• 💡 Not just technical ➔ communication and curiosity rank alongside statistics and programming.

🧠 Active Recall

Distinguish data analyst, data scientist, and data engineer by their primary output.

Answer

• Short answer: Analyst → insights from data; scientist → data models/products that generate insights; engineer → data infrastructure, automated processing, model deployment at scale.
• Why: “Primarily” ➔ boundaries blur — a data scientist out-stats an engineer and out-engineers a statistician.